一种自信学习自动机  

A Self-Confident Learning Automaton

作　　者：刘晓[1] 

机构地区：[1]中国航空计算技术研究所,西安710068

出　　处：《航空计算技术》1998年第1期13-15,共3页Aeronautical Computing Technique

摘　　要：提出一种二次不动—惩罚变结构随机自动机模型（Q(IP)）。较之于其线性形式（L(IP)），新模型的学习带有一定的自信（当然，有时也可能是自负）。特别，跟传统自动机不同的是，新算法的极限行为同时兼具吸收壁和遍历性。In this paper, a model of the quadratic inaction-penalty variable structure stochastic automata, QIP, is presented. Compared with its linear counterpart (Lin),learning of the quadratic automaton is self-confident (and may also be, of cause, self-opinionated in some extremely bad cases). Especially, as opposed to the traditional automata, the limiting behavior of the proposed algorithm possesses both absorbing barriers and ergodicity.

关 键 词：学习自动机 变结构 随机自动机 强化学习 

分 类 号：TP391.6[自动化与计算机技术—计算机应用技术]